Optimisation et étalonnage d’un Robot Sériel pour l’Impression 3D

Abstract

This thesis presents an approach to extend the capabilities of vat photopolymerization (VPP) 3D printing using a serial robotic arm. The study is structured in two parts. The first introduces a semi-automated, low-cost, and portable calibration method. A TriCal probe (three orthogonal dial indicators) is mounted on the resin vat through a kinematic coupling that can be repositioned in three repeatable orientations. A calibration tool carried by the robot holds three precision spheres serving as measurement references. Through iterative automatic recentering, the robot successively aligns each sphere with the probe origin across a carefully chosen set of joint–rail postures. The geometric model relies on the modified Denavit–Hartenberg convention and incorporates a joint compliance term to account for gravity effects on the robot. From roughly 3,300 measurable configurations in simulation, 40 poses are selected using observability analysis (index ▯1) and a DetMax algorithm, then sequenced via a Traveling Salesman Problem solver and collision-free planning. Iterative parameter identification (linearized least squares) covers 34 parameters (arm, rail, vat pose). Independent validation with a laser tracker shows an improvement in relative accuracy in the vat frame from 1.272 mm to 0.271 mm on random poses, and to 0.220 mm on average along a representative printing path. The second part addresses optimization of the printing stage. The kinematic chain is inverted : the robot flange is fixed to the world, and the “TCP” corresponds to the polymerization zone at the vat bottom. A virtual degree of freedom for screen rotation (SLA) is introduced, bringing the system to 8 DOFs. Inverse kinematics is solved under constraints (pose tolerance, joint limits, collision avoidance, continuity) and evaluated through multi-objective metrics : translational and rotational manipulability, normalized joint-limit margin, and positional/angular sensitivity. The algorithm follows a deterministic coarse-to-fine strategy : discrete sweeping of global part orientations, initial sampling via Halton sequences, feasibility filtering on a sparsified trajectory, then refinement and multi-criteria ranking. This pipeline rapidly eliminates infeasible orientations and selects printing trajectories with wide joint margins and low error amplification at the contact point. The main contributions are : a calibration method tailored to robotized VPP and validated experimentally; a reproducible multi-objective optimization framework consistent with a calibrated model ; and concrete perspectives (enhanced VPP capabilities, redundancy resolution favoring the rail axis, periodic recalibration) toward industrial deployment.

Date	14 Nov 2025
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Ilian Bonev (Supervisor)